Abstract:
An efficient implementation of a Distributed Speech Recognition (DSR) on a Digital Signal Processor (DSP) platform is shown. Although complying with the European Telecommunications Institute (ETSI) DSR standard, feature extraction and data compression enhancements are added, in order to address three issues: computational speed of the signal analysis process, reliability of the representative signal features, and low channel bandwidth utilization, when transmitting the speech features. A series of tests and simulations are provided, in order to show the functionality of the proposed system.